CRISPR Scoring

SnapGene 9.0 and later provides users with the ability to score their 5'-N20-NGG Cas9 CRISPR sites using a number of published activity and specificity algorithms. Scoring requires a named-user license and an Internet connection to access our secure cloud server. Prior to the SnapGene 9.0 release, users will need to request cloud access, for more information see – Accessing CRISPR Cloud Scoring. An outline of each algorithm is given below:

On Target Efficiency Scoring

On-target (or activity) scoring looks at the sequence features of the CRISPR site itself, and compares them to an experimentally determined model. The model then scores the site based on its predicted level of activity. SnapGene 9.0 offers activity scoring for Cas9 CRISPR sites using the methods from Doench et al. (2014), Doench et al. (2016) and DeWeirdt et al. (2022). 

Doench 2014 - Doench et al. 2014

Also known as Rule Set 1, and has been trained on over 1,800 sgRNAs. This method analyzes the one- and two-base features of the gRNA, as well as the GC content, to generate a score between 0 and 1. Higher scores indicate a better predicted on-target activity for that guide. Note that CRISPR sites with ambiguous bases will have an undefined score.

Doench 2016/Azimuth - Doench et al. 2016

Also known as Rule Set 2, the Azimuth, or Fusi-score. An improved regression model trained on over 4,000 sgRNAs, that also accounts for the sgRNA location within a protein sequence, melting temperatures of features of the sgRNA and the sgRNA sequence as a whole, and counts of single and dinucleotides in a position-independent manner. Scores are between 0 and 1, with a higher score denoting higher expected activity. CRISPR sites with ambiguous bases will have an undefined score. 

Doench 2022/RS3 - DeWeirdt et al. 2022

Also know as Rule Set 3. Doench 2022 gives a prediction of the on-target activity of a guide, trained on over 47,000 sgRNAS, and accounting for tracrRNA variants when scoring sgRNA activity. SnapGene 9.0 supports scoring with the Hsu et al. (2013) tracrRNA variant (tracrRNA starting with GTTTTAGAGCTA...). This tracrRNA variant will be appropriate for most use cases, and is used by most on-target scoring models. Scores for Doench 2022 are between -3 and +2 with a higher score denoting higher expected activity.

Off-Target Specificity Scoring

SnapGene 9.0 offers the CFD (Doench et al. 2016) and MIT (Hsu et al. 2013) specificity scoring algorithms. These algorithms compare each CRISPR site to a specified genome and score based on how unique the site is, and how likely it will cause off-target effects. You can choose from a set list of preindexed genomes, see table below.

CFD specificity score - Doench et al. 2016

The cutting frequency determination (CFD) score provides a measure of the likelihood of Cas9 cleavage occurring at off-target sites (Doench et al, 2016). Each predicted off-target site is assigned an individual score, which is then aggregated to give an overall guide-level specificity score between 0 and 100 - the higher the aggregated score, the lower the predicted likelihood of off-target cleavage. SnapGene allows up to 3 maximum mismatches when calculating these scores.

The CFD score is considered a more accurate measure of guide specificity than the MIT score, as it was derived empirically from systematic mismatch tolerance data, accounting for both the position and nature of each mismatch along the guide sequence (Doench et al., 2016).

MIT specificity score - Hsu et al. 2013

The MIT specificity score, also known as the Hsu or Zhang score, predicts the likelihood of Cas9 cleavage occurring at off-target sites (Hsu et al., 2013). Each predicted off-target site is assigned an individual score based on the number and position of mismatches relative to the guide sequence, which are then aggregated to give an overall guide specificity score between 0 and 100 - higher values indicate a lower predicted likelihood of off-target cleavage. SnapGene allows up to 3 maximum mismatches when calculating these scores.